How to install torchsparse in Windows10?

[Huhaowen0130]

No description provided.

[kentang-mit]

Hi,

I haven't tried the installation on Windows 10. If you encounter any problem when installing torchsparse on Windows10 please reply me in this issue and I will try to figure out the solution.

Best,
Haotian

[zhijian-liu]

I'm closing this issue due to the inactivity. Please feel free to reopen it if the problem has not been resolved.

[weizmanl]

I have Windows 10 and I have been trying to install torchsparse, but no success.

When I run pip things compile, but linkage fails. It says:
Creating library C:\Users\WeizmanL\AppData\Local\Temp\pip-req-build-6qzmrtnx\build\temp.win-amd64-3.8\Release\torchsparse/src\torchsparse_cuda.cp38-win_amd64.lib and object C:\Users\WeizmanL\AppData\Local\Temp\pip-req-build-6qzmrtnx\build\temp.win-amd64-3.8\Release\torchsparse/src\torchsparse_cuda.cp38-win_amd64.exp
hash_cpu.obj : error LNK2001: unresolved external symbol "public: long * __cdecl at::Tensor::data_ptr(void)const " (??$data_ptr@J@Tensor@at@@QEBAPEAJXZ)
build\lib.win-amd64-3.8\torchsparse_cuda.cp38-win_amd64.pyd : fatal error LNK1120: 1 unresolved externals
error: command 'C:\Program Files (x86)\Microsoft Visual Studio\2019\BuildTools\VC\Tools\MSVC\14.28.29333\bin\HostX86\x64\link.exe' failed with exit status 1120

Any ideas?

Thank you

[Huhaowen0130]

I have Windows 10 and I have been trying to install torchsparse, but no success.

When I run pip things compile, but linkage fails. It says:
Creating library C:\Users\WeizmanL\AppData\Local\Temp\pip-req-build-6qzmrtnx\build\temp.win-amd64-3.8\Release\torchsparse/src\torchsparse_cuda.cp38-win_amd64.lib and object C:\Users\WeizmanL\AppData\Local\Temp\pip-req-build-6qzmrtnx\build\temp.win-amd64-3.8\Release\torchsparse/src\torchsparse_cuda.cp38-win_amd64.exp
hash_cpu.obj : error LNK2001: unresolved external symbol "public: long * __cdecl at::Tensor::data_ptr(void)const " (??$data_ptr@J@Tensor@at@@QEBAPEAJXZ)
build\lib.win-amd64-3.8\torchsparse_cuda.cp38-win_amd64.pyd : fatal error LNK1120: 1 unresolved externals
error: command 'C:\Program Files (x86)\Microsoft Visual Studio\2019\BuildTools\VC\Tools\MSVC\14.28.29333\bin\HostX86\x64\link.exe' failed with exit status 1120

Any ideas?

Thank you

Well, I have tried some solutions about the installation and failed, perhaps it's better to install on linux.

[carlosedubarreto]

I was able to install making some changes.

First of all, dont do that "ninja" fix. the one that you change "ninja -v" to "ninja --version". This will make your compiling to have new errors that should not be there (like the difficult to compile files that it already have access, I'm saying it because it happened to me 😅)

if you didnt change it, just go to the next step.

I noticed that it was complaining that it was missing some sparsehash include code. ("google/dense_hash_map")
I went to
https://github.com/sparsehash/sparsehash
and got the version 2.0.4

https://github.com/sparsehash/sparsehash/releases/tag/sparsehash-2.0.4

after getting this file, you will have a zip file.
Get the contents that you have in the src folder

and paste it on you python enviroment include folder.
in my case it was

C:\Users\Pichau.conda\envs\sparseneus\include

it was like this after i pasted them

This will solve the issue of
"google/dense_hash_map"

the other issue I had was related to this error
"public: long * __cdecl at::Tensor::data_ptr(void)const " (??$data_ptr@J@Tensor@at@@QEBAPEAJXZ)

To solve it I saw what was commented on these two places

https://forums.developer.nvidia.com/t/strange-link-error-seen-by-multiple-people-while-building-pytorch-cpp-cuda-extensions/145261

and

isl-org/StableViewSynthesis#15

They say that there is a problem when we try to build on windows, about some type of variables, to change from "long" to "int64_t".

I was able to mae it work changeing these files:
torchsparse\backend\others\query_cuda.cu
torchsparse\backend\others\query_cpu.cpp

torchsparse\backend\hash\hash_cuda.cu
torchsparse\backend\hash\hash_cpu.cpp

to run the installation I ran "python setup.py install"
and it successfully built and installed torchsparse on my venv

For people that dont want to change the files from torchsparse by itself, I made a fork and updated those files.
Probably it will be easier to use this repository I forked
https://github.com/carlosedubarreto/torchsparse

Hope it helps

[denred0]

I was able to install making some changes.

First of all, dont do that "ninja" fix. the one that you change "ninja -v" to "ninja --version". This will make your compiling to have new errors that should not be there (like the difficult to compile files that it already have access, I'm saying it because it happened to me 😅)

if you didnt change it, just go to the next step.

I noticed that it was complaining that it was missing some sparsehash include code. ("google/dense_hash_map") I went to https://github.com/sparsehash/sparsehash and got the version 2.0.4

https://github.com/sparsehash/sparsehash/releases/tag/sparsehash-2.0.4

after getting this file, you will have a zip file. Get the contents that you have in the src folder

and paste it on you python enviroment include folder. in my case it was

C:\Users\Pichau.conda\envs\sparseneus\include

it was like this after i pasted them

This will solve the issue of "google/dense_hash_map"

the other issue I had was related to this error "public: long * __cdecl at::Tensor::data_ptr(void)const " (??$data_ptr@J@Tensor@at@@QEBAPEAJXZ)

To solve it I saw what was commented on these two places

https://forums.developer.nvidia.com/t/strange-link-error-seen-by-multiple-people-while-building-pytorch-cpp-cuda-extensions/145261

and

isl-org/StableViewSynthesis#15

They say that there is a problem when we try to build on windows, about some type of variables, to change from "long" to "int64_t".

I was able to mae it work changeing these files: torchsparse\backend\others\query_cuda.cu torchsparse\backend\others\query_cpu.cpp

torchsparse\backend\hash\hash_cuda.cu torchsparse\backend\hash\hash_cpu.cpp

to run the installation I ran "python setup.py install" and it successfully built and installed torchsparse on my venv

For people that dont want to change the files from torchsparse by itself, I made a fork and updated those files. Probably it will be easier to use this repository I forked https://github.com/carlosedubarreto/torchsparse

Hope it helps

Thanks man! You did my day!

[initialneil]

For anyone reach this issue after 2022.12. With cuda 11.3 or newer maybe, some additional changes are needed for the above post.

1. Run vcvars64.bat first to enable Visual Studio (to enable cl.exe).
2. Files in windows also need to be copied to envs's include folder (fix the error of cannot find sparseconfig.h).
3. Comment out the function in torchsparse\backend\utils\atomic.cuh (fix the error of atomicExch already defined).

[tiancaipipi110]

I did everything @denred0 and @initialneil suggested, and got two other errors:

1. ImportError: cannot import name '_nt_quote_args' from 'distutils.spawn'
2. subprocess.CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 1

To fix these two errors, I downgraded setuptools by conda install -c conda-forge setuptools=59.6 suggested here and changed 'command = ['ninja', '-v']' to 'command = ['ninja', '--version']' in cpp_extension.py suggested in this post

But now I get a LINK : fatal error LNK1181: cannot open input file

OS: Win10
python 3.7.16
torchvision 0.7.0
pytorch 1.6.0
cudatoolkit 10.2.89
cuda 11.7

Here's my vs BuildTools

running install
C:\Users\junew\anaconda3\envs\leRes\lib\site-packages\setuptools\command\install.py:37: SetuptoolsDeprecationWarning: setup.py install is deprecated. Use build and pip and other standards-based tools.
  setuptools.SetuptoolsDeprecationWarning,
C:\Users\junew\anaconda3\envs\leRes\lib\site-packages\setuptools\command\easy_install.py:159: EasyInstallDeprecationWarning: easy_install command is deprecated. Use build and pip and other standards-based tools.
  EasyInstallDeprecationWarning,
running bdist_egg
running egg_info
writing torchsparse.egg-info\PKG-INFO
writing dependency_links to torchsparse.egg-info\dependency_links.txt
writing top-level names to torchsparse.egg-info\top_level.txt
reading manifest file 'torchsparse.egg-info\SOURCES.txt'
adding license file 'LICENSE'
writing manifest file 'torchsparse.egg-info\SOURCES.txt'
installing library code to build\bdist.win-amd64\egg
running install_lib
running build_py
running build_ext
C:\Users\junew\anaconda3\envs\leRes\lib\site-packages\torch\utils\cpp_extension.py:270: UserWarning: Error checking compiler version for cl: [WinError 2] The system cannot find the file specified.
  warnings.warn('Error checking compiler version for {}: {}'.format(compiler, error))
building 'torchsparse.backend' extension
Emitting ninja build file C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\build.ninja...
Compiling objects...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
1.10.2
C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.34.31933\bin\HostX86\x64\link.exe /nologo /INCREMENTAL:NO /LTCG /DLL /MANIFEST:EMBED,ID=2 /MANIFESTUAC:NO /LIBPATH:C:\Users\junew\anaconda3\envs\leRes\lib\site-packages\torch\lib /LIBPATH:C:\Users\junew\anaconda3\envs\leRes\libs /LIBPATH:C:\Users\junew\anaconda3\envs\leRes\PCbuild\amd64 "/LIBPATH:C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.34.31933\lib\x64" "/LIBPATH:C:\Program Files (x86)\Windows Kits\NETFXSDK\4.8\lib\um\x64" "/LIBPATH:C:\Program Files (x86)\Windows Kits\10\lib\10.0.20348.0\ucrt\x64" "/LIBPATH:C:\Program Files (x86)\Windows Kits\10\\lib\10.0.20348.0\\um\x64" c10.lib torch.lib torch_cpu.lib torch_python.lib /EXPORT:PyInit_backend C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\pybind_cpu.obj C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\convolution\convolution_cpu.obj C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\devoxelize\devoxelize_cpu.obj C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\hash\hash_cpu.obj C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\hashmap\hashmap_cpu.obj C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\others\count_cpu.obj C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\others\query_cpu.obj C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\voxelize\voxelize_cpu.obj /OUT:build\lib.win-amd64-3.7\torchsparse\backend.cp37-win_amd64.pyd /IMPLIB:C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\backend.cp37-win_amd64.lib
LINK : fatal error LNK1181: cannot open input file 'C:\Users\junew\anaconda3\envs\LeReS\Lib\site-packages\torchsparse-1.4.0\build\temp.win-amd64-3.7\Release\torchsparse\backend\hashmap\hashmap_cpu.obj'
error: command 'C:\\Program Files (x86)\\Microsoft Visual Studio\\2022\\BuildTools\\VC\\Tools\\MSVC\\14.34.31933\\bin\\HostX86\\x64\\link.exe' failed with exit status 1181

[limaodaxia]

They say that there is a problem when we try to build on windows, about some type of variables, to change from "long" to "int64_t".

Hello, why change "long" to "int64_t" ? "int64_t" seems not to be a legal datatype? I change "long" to "long long" and it worked.